I. Core Research Summary

• Fully customizable 3D probe architecture

CAD‑based digital design allows for arbitrary probe height, diameter, and arrangement; the entire device is fabricated in a single step. It can produce micropillars ranging from 0.1 to 3 mm in length, with tip diameters as small as 10 μm and aspect ratios up to 50:1. The system supports square and hexagonal high‑density arrays, with hexagonal arrangements achieving a 15% increase in area density and a maximum density of 2,600 shanks/cm² (at 200 μm pitch).

• Multilayer aerosol wiring system

Aerosol jet printing alternates between depositing metallic conductive layers and polyimide insulating layers, enabling ultra-thin multilayer signal traces with an overall trace thickness of only around 10 μm—significantly thinner than conventional PCBs fabricated via photolithography. High‑density probes can also be routed through independent channels, eliminating crosstalk.

• Excellent mechanical and electrical properties

Metal microcolumns exhibit exceptionally high buckling resistance, with a puncture force only 1/80th of the buckling failure load. After PEDOT:PSS coating, the probe’s impedance at 1 kHz drops to as low as 237 kΩ; among 230 channels, the yield rate exceeds 98.8%, and after 15 repeated penetrations into brain tissue, the coating remains intact with stable impedance.

• Minimally Invasive In Vivo Neural Recording

The ultrafine conical probe can penetrate the mouse brain without causing extensive tissue tearing, enabling simultaneous recording of neuronal action potentials from multiple cortical and hippocampal regions, with an average signal-to-noise ratio of 8.6—comparable to that of commercial silicon probes.

• Ultra-fast, low-cost manufacturing

Traditional silicon photolithography array fabrication takes several weeks, whereas aerosol printing can produce a complete device in just a few hours; silver-based inks significantly reduce material costs, and the process is compatible with multiple substrates, including flexible PI, alumina, and PCBs.

II. Core Working Principle and Unique Process Advantages of Aerosol-Based 3D Nanoprinting

(1) Complete Integrated Printing Workflow

• Ultrasonic atomization of gold–silver nano‑ink and pneumatic atomization of polyimide insulating ink generate uniform aerosol droplets with diameters of 1–5 μm.

2. Nitrogen carrier gas delivery, with a precisely focused coaxial sheath gas jet, enabling non-contact deposition from the nozzle at distances of 1–5 mm;

• The platform precisely positions and sequentially stacks metal nanoparticles, with the solvent evaporating in situ to induce solidification, enabling the layer-by-layer growth of three-dimensional microprobes with high aspect ratios.
• Alternate printing of conductive traces and insulating dielectric layers to complete multilayer signal routing.
• Vacuum‑deposited parylene provides uniform insulating encapsulation, while a PFIB ion beam selectively exposes the probe tip; electrochemical deposition of PEDOT:PSS reduces interfacial impedance.

(II) Five Core Process Advantages (Compared with Silicon Photolithography/Etching and Utah Arrays)

• Fully digital, freely customizable, with no mold constraints.

No etching, cutting, or masking is required; CAD enables one-click adjustment of probe length, spacing, and layout, allowing for the creation of region-specific arrays tailored to different brain areas and accommodating diverse animal‑experiment and clinical customization needs.

• Ultra-high channel density, breaking through the bottleneck of silicon‑process node pitch.

The conventional silicon sawing process achieves a minimum pitch of 400 μm, while aerosol printing can achieve as low as 100 μm. Hexagonal array patterning further increases the number of probes per unit area, enabling high‑precision, large‑area neural signal recording.

• Flexible / Rigid Full-Substrate Compatibility

Alumina ceramics, PCBs, and polyimide flexible films can all be directly printed, and the flexible substrate array is well-suited for sensing curved biological tissues such as the heart.

• Three-dimensional, one-step molding eliminates the assembly process.

Traditional probes require wafer dicing and manual assembly of multi-layer circuits, whereas aerosol‑assisted simultaneous printing enables the fabrication of three‑dimensional electrodes and underlying interconnects in a single step, eliminating interlayer delamination defects.

• Short-cycle, low-cost additive manufacturing

The fabrication of a single batch of devices has been shortened from several weeks to just a few hours; silver nanoparticle ink replaces costly silicon wafers and gold‑plated wafers, with extremely low material waste, significantly lowering the barrier to entry for brain–computer interface device development.

III. Five Core Technological Innovations of This Study

Innovation 1: A novel process for layer-by-layer aerosol‑based 3D microcolumn fabrication

Innovation 2: Aerosol Multi-Layer Metal–Insulator Composite Wiring Technology

Innovation: A Multi-Material Aerosol-Compatible System

Innovative 4-cone microprobe minimally invasive structural design

Innovative, end-to-end post-processing support processes

IV. Existing Process Challenges and Corresponding Solutions for Overcoming Them

Challenge 1: Aerosol printing is prone to edge fogging (overspray), and electrical leakage can cause short circuits between probes.

Challenge 2: Metal microcolumns with a high aspect ratio are prone to bending and fracture under stress.

Challenge 3: A single ink cannot simultaneously achieve both three-dimensional probing and fine wiring.

Challenge 4: Conventional conductive probes exhibit high interfacial impedance, resulting in a poor signal-to-noise ratio for neural signals.

Challenge 5: Traditional techniques can only standardize probe height, failing to accommodate the brain’s layered architecture.

V. Future Diversified Industrial Application Scenarios

1. Research at the Cutting-Edge Neuroscience Laboratory

2. Minimally Invasive Brain-Computer Interface Medical Device

3. Organ-on-a-chip biosensing

4. Long-term recording device for animal behavior studies

5. Flexible Implantable Bioelectronics

Figure 1. Example application scenarios and the printing workflow. (A) To understand the interactions among neural circuits, it is necessary to record activity from multiple brain regions. The specific target areas may vary depending on the study. In this conceptual diagram, the targeted regions—listed from left to right—are the mouse prefrontal cortex, motor cortex, caudal striatum, and hippocampal CA2 and CA3 areas. (B) A real mouse brain slice illustrates the targeted localization of the regions described in (A). A sagittal section shows multiple 3D‑printed probe shafts (in red), each penetrating the cortex, striatum, and hippocampus; these penetration paths are arranged in parallel. The visibility of the shaft trajectories results from dehydration of the tissue during histological processing. (C) AJ Company employs 3D printing to rapidly fabricate a robust, fully customizable neural probe design, incorporating a circuit board for transmitting physiological signals. (D) An illustrative image of the 32‑channel probe during the printing process. For detailed information on the printing workflow, please refer to Videos S1, S2, and S3. (E) A schematic of the 32‑channel probe design, including signal transmission pathways and connectors. (F) A photograph of the printed probe based on the design shown in (E), with a U.S. penny used as a scale reference.

Table 1. Manufacturing capabilities and current limitations of the 3D‑printed CMU array described in this paper. Processing capabilities and limitations of the 3D nanofabrication process employed for the neural probe developed in this study. PET: polyethylene terephthalate; PTFE: polytetrafluoroethylene.

Figure 2. Fabrication of CMU arrays using three-dimensional nanoparticle printing technology. (A) A rod‑shaped array comprising 512 electrodes, with adjustable rod heights, demonstrating the method’s potential for customization. The rod lengths follow a stepped design to ensure regularity; each length can be independently set to any desired value. A time‑lapse video of the probe assembly process is provided in Movie S2. (B) A printing platform employing 1‑mm‑long rods with an aspect ratio of 1:50, fabricated on an alumina ceramic substrate. Using this approach, 1‑mm‑long rods can be produced in under 1.5 minutes. (C) The printed 10×10 array has a height of 2.4 mm and an average width of 30 μm; in plan view, the rods appear straight and uniform, with the entire structure designed according to a consistent geometric motif. (D) Close‑up of a conical rod tip, facilitating insertion into tissue. (E) An example of rod‑shaped structures directly printed onto a flexible polyimide (Kapton) substrate; integrated conductive traces on the same substrate enable connection to data‑acquisition hardware. (F) An example of rod‑shaped structures printed using gold nanoparticles; here, a wider base is employed to enhance stability. Note that, to highlight different geometries, the other panels in this figure were printed with lower‑cost silver ink. (G) Rod‑shaped structures can be customized in arbitrary configurations; the rods shown here mimic the shape of the Carnegie Mellon University acronym “CMU.” (H) A high‑density, high‑aspect‑ratio array is printed in a hexagonal stacking pattern, with a pitch of 200 μm (plan view). By arranging the electrodes in a hexagonal rather than square configuration, the areal electrode density can be increased by 15%.

Figure 3. Electrical wiring architecture of the high-density probe array. A custom circuit board is fabricated using multilayer metal–polymer printing technology to route signals from a high-density microelectrode array (MEA). (A) Schematic illustrating the step-by-step layer-by-layer printing process of the multilayer board, presented in an exploded view; (B) Optical image of the printed circuit board. The layers are labeled sequentially from bottom to top as (i) through (vi): (i) the first metal layer is printed on an alumina ceramic substrate, with pads at both ends of each trace; (ii) a polyimide layer is printed to insulate the metal traces, except at the pads; (iii) an intermediate polyimide layer separates the fourth layer from the first; (iv) the upper metal layer is printed to create two conductive layers for signal transmission; (v) a polymer insulating layer is printed over the surface of the upper metal layer, excluding the pad regions at both ends. Note that by simply modifying the printing sequence, the signal routing paths can be customized, enabling arbitrary placement and elevation of probes and pads; (vi) probe posts are printed onto the exposed pads using an AJ printing process, ultimately yielding a 2 mm × 2 mm circuit board containing 100 probe posts.

Figure 4. Functionalized architecture of the 3D‑printed array. (A and B) Schematic illustration and actual photographs of the solder deposition process for connecting a standard connector to the 3D‑printed probe; (C) Photograph of the joint after reflow soldering; (D and E) Schematic diagram and scanning electron microscope images showing the removal of overhang layers via PFIB etching, with black regions indicating substrate‑visible areas (i.e., where the overhang layer has been eliminated or was never present); (F) After Parylene‑C coating, PFIB processing is used to reveal the probe tip, enabling the probe to record neural voltage signals only from the exposed region; (G and H) Scanning electron microscope images before and after selective electrodeposition of PEDOT:PSS onto the probe tip, demonstrating the impedance‑reducing effect.

Figure 5. Electrical, electrochemical, and mechanical properties of 3D‑printed probes. (A) Resistivity of printed gold and silver devices at different sintering temperatures. Under the sintering conditions employed in this study—corresponding to the glass transition temperature (GTT)—the resistivities differ from those of bulk metals by one to two orders of magnitude. NP: nanoparticles. (B) Electrical impedance measurements at 1 kHz for 230 probe shafts across nine devices. All 230 channels exhibit electrical isolation after fabrication, with impedances below 1 MΩ at 1 kHz. (C and D) Impedance data for gold and silver electrodes, respectively, with and without a PEDOT:PSS coating; error bars represent the standard deviation across five independent channels. In all cases, the PEDOT:PSS coating reduces the impedance at 1 kHz to approximately 100–250 kΩ. Individual impedance traces for gold and silver probe shafts with and without the PEDOT:PSS coating are shown in Figure S5 (panels A–E and F–J, respectively). (E) Peak insertion forces measured when probes of varying lengths are inserted into an agar‑based brain model and compressed by a solid platen; in each case, the insertion forces are several orders of magnitude lower than the compressive forces required to induce bending. (F) Effect of repeated acute insertions of silver probe shafts into the brain model (agar): blue data points show the average of three replicate measurements after a given number of insertions. The figure also displays the mean and standard deviation of the impedance values corresponding to the blue data points. (G) Representative scanning electron microscope images of the shank ends shown in (F), before insertion and after 3×, 8×, and 15× insertions. Both the PEDOT:PSS coating and the poly(p‑xylylene)-C coating remain intact after 15 insertions, consistent with the impedance measurements reported in (F). Scale bar: 10 μm.

Figure 6. Procedure for inserting the probe into the mouse brain. (A and B) Schematic diagrams and photographs illustrating the structures targeted during recording with a standard stereotaxic micromanipulator. (C) Evans blue–stained traces left after successful insertion of a probe consisting of a 10×10 array of densely packed fibers (density: 2,600 fibers/cm²; spacing: 200 μm), with a tip diameter of 20 μm. (D) No obvious tissue damage was observed after 30 minutes of insertion [blue: 4′,6-diamidino‑2‑phenylindole (DAPI) nuclear staining; red: Evans blue (EB)]. (E) Successful probe insertion in an anesthetized mouse without probe breakage or evident brain injury (coronal section showing area V2 and the hippocampus). Note that the absence of damage or tearing is attributable to the probe‑insertion protocol described in (C) through (E).

Figure 7. Schematic illustration of action potential recordings from the mouse brain. (A) Spike waveforms extracted from a 32-channel electrode implanted in the mouse sensorimotor cortex. Scale bar: 1 ms × 100 μV; different colors denote independently isolated neurons (black: neuron 1; red: neuron 2; blue: neuron 3; green: neuron 4). (B) Enlarged views of randomly selected waveforms originating from one-, two-, or three‑channel segments presumed to contain single neurons, with the same scale as (A). (C and D) Waveforms obtained after a second implantation of a different electrode into the sensorimotor cortex, using the same scale as (A) and (B). (E) Typical trace of single‑channel neural activity, with a time scale of 100 ms. (F) and (G) show, respectively, the identified neurons in (A) (magenta), the identified neurons in (B) (cyan), and the distribution of mean firing rates and signal‑to‑noise ratios across all eight recording sessions (black).